Generiranje umjetnih slika zuba pomoću dubokih neuronskih mreža

master's thesis

• Title (english): Generating Artificial Images of Teeth Using Deep Neural Networks
• Author: Marija Batur
• Mentor: Marko Subašić (mentor)
• Committee member: Marko Subašić (predsjednik povjerenstva)
• Committee member: Juraj Petrović (član povjerenstva)
• Committee member: Krešimir Križanović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Rad se bavi generativnim modelima za stvaranje umjetnih slika, konkretno slike pojedinih zuba korištenjem generativne suparničke mreže. Opisani su dobiveni podatci i način na koji se prilagođavaju za treniranje. Dan je pregled dubokih generativnih modela, usredotočen na modele koji se koriste za generaciju slika. Detaljnjije je proučen način rada generativnih suparničkih mreža, njihov različiti modeli i područja uporabe. Dan je detaljan prikaz implementiranog modela, u biblioteci PyTorch,... More temeljen na arhitekturi DCGAN, kao i procesa treniranja i dodataka za poboljšanje. Objašnjeni su načini evaluacije dobivenih rezultata. Predstavljeni su rezultati implementirane mreže, te rezultati eksperimenata s korištenjem gubitka klasifikatora pri treniranju i dodatkom gradijenta diskriminatora generiranim slikama. Less

Abstract (english)

This paper examines generative models for artificial image synthesis, specifically images of individual teeth using generative adversarial networks. Available data and its preparation for training are described. An overview of deep generative models is given, focused on image-generating models. Generative adversarial networks are studied in more detail, with their main concepts, different models and uses. The implemented model is described, implemented using PyTorch library, based on... More DCGAN architecture, with the description of the training process and additional improvements. Evaluation metrics for achieved results are explained. The results obtained by the implemented network are presented, including experiments with the use of classifier loss in generator training and adding discriminator gradient to the generated images.
Less

• Keywords: duboko učenje; generativni modeli; generativne suparničke mreže
• Keywords (english): deep learning; generative models; generative adversarial networks
• Language: croatian
• URN:NBN: urn:nbn:hr:168:466874
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 14:31:14